rds_exporter breaks when some instances don't have enhanced monitoring enabled
Description
How to test
How to document
Attachments
Confluence content
Activity
Ivan Groenewold December 22, 2021 at 12:31 PM
hmm... the only thing I can think of that seems to be different from my test case is I was using IAM role instead of access keys. Can you try that?
Jiří Čtvrtka December 22, 2021 at 12:27 PM(edited)
Checked with auroras. Take a look if I didnt miss something. Thank you.
Ivan Groenewold December 21, 2021 at 11:43 AM
Hi, I see you are testing with Postgres instances. Can you try reproducing by following the steps as listed in the description but with MySQL instances? Also take a look at the related ticket where we fixed this for PMM 1.x I am sure the issue is the same.
Jiří Čtvrtka December 21, 2021 at 11:36 AM
For instance with enhanced metrics it is ok. Disable one is not even in dropdown. So this is correct right?
Roma Novikov December 21, 2021 at 8:33 AM
I recommend checking "do we have metrics" not on the DB dashboard but in VM or metrics themselves (ex: Advanced Data exploration ) because DB dashboards are designed to extract data from different available sources.
Our RDS exporter exposes several metrics, some of which are about the same part of the system.
Enhanced monitoring is more preferable way as it's consumes less API calls the basis one (expectations)
Details
Details
Assignee
Reporter
Priority
High
Impact on the user: Addin only one RDS instance w/o enhanced monitoring is stopping PMM from collecting OS-level data for all instances. No data for RDS with enhanced monitoring on
There are issues when mixing instances with and without enhanced monitoring to PMM. After adding any instance without enhanced monitoring, the OS-level metrics stop being collected for ALL existing instances.
Steps to reproduce:
Add RDS/Aurora MySQL instance (instance1) configured with enhanced monitoring=on
Check Node Summary for instance1, all good
Add another RDS/Aurora MySQL instance (instance2) with enhanced monitoring=off
Actual result:
Check Node Summary for instance1, no metrics reported anymore
Expected result:
Node metrics are still present for instances where enhanced metrics ON
Workaround:
Turn on Enhanced monitoring to all RDS instances. Note: this will cost $, so not a good workaround if the user needs to save some money